Stephanie Kollmann 1, Geraldine Celliere 1, Monika Twarogowska 1
1 Simulations Plus (Paris, France)
INTRODUCTION
The FDA’s October 2025 draft update to Scientific Considerations in Demonstrating Biosimilarity marks a significant evolution in biosimilar development. Rather than expecting confirmatory clinical trials by default, the updated framework links the need for additional studies to the residual uncertainty remaining after analytical and PK/PD evaluation. When this uncertainty is sufficiently addressed, further comparative efficacy studies (CES) may no longer be required. This shift increases the strategic importance of mechanistic PK/PD modeling and places it at the center of biosimilar study design.
For pharmacometricians, it creates a clear opportunity: modeling can directly inform dose selection, sensitivity assessment, and sample-size justification in biosimilarity programs. However, practical frameworks to translate this regulatory flexibility into concrete design decisions remain limited [1]. Using pegfilgrastim, a biologic with nonlinear target-mediated PK/PD, as a case study, we developed a reproducible, single-environment (MonolixSuite with R automation) workflow to (i) identify the dose most sensitive to plausible product differences and (ii) optimize study size under realistic within-subject variability. Openly available templates facilitate application across biosimilar development programs.
METHODS
The analysis followed a two-step model-informed framework aligned with regulatory decision-making.
Step 1 – Parameter sensitivity and dose selection:
A published mechanistic population PK/PD model for pegfilgrastim (quasi-equilibrium TMDD with ANC feedback) was used to quantify the sensitivity of standard similarity endpoints (AUC, Cmax, ANC AUEC, ANCmax) across 2, 4, and 6 mg doses. Controlled perturbations were introduced to three biosimilarity-relevant drivers: delivered dose (Fsc), potency (receptor-binding affinity Kd), and linear clearance (CL), representing plausible product-related differences. The objective was to identify the dose most sensitive to potential differences in PK/PD metrics between biosimilar and reference [2].
Step 2 – Study design optimization:
Clinical trial simulations replicated a typical 2×2×2 crossover PK/PD biosimilarity study (RT/TR sequences; single 6 mg per period). Replicates were simulated across sample sizes (50–250 subjects) and parameter-difference scenarios. For each replicate, bioequivalance testing was performed, and empirical power of 80% was calculated as the proportion of replicates concluding biosimilarity [3]. It allowed identification of minimum sample size required under varying parameter perturbations (Fsc, Kd, Cl) and assumptions of inter-occasion variability (15–20%).
All simulations and analyses were implemented within a single MonolixSuite–R automated workflow to ensure reproducibility and transferability.
RESULTS
Parameter sensitivity analysis demonstrated that PK exposure metrics (AUC, Cmax) were sensitive to controlled perturbations in delivered dose, potency, and clearance, whereas PD endpoints (ANC AUEC, ANCmax) remained largely buffered due to feedback regulation. Sensitivity patterns differed by dose: lower doses were more responsive to potency differences, while at 6 mg (approved therapeutic dose) exposure became more sensitive to clearance-related changes as target-mediated processes approached saturation. These results illustrate how mechanistic PK/PD modeling can identify the dose that is the most sensitive for addressing potential product-differences. For subsequent powering simulations we used 6 mg.
Clinical trial simulations showed that study power was driven primarily by PK endpoints. Under realistic variability (IOV = 20%), approximately 200 subjects were required to achieve ~80% power for modest product-related differences, while larger perturbations rapidly reduced the probability of concluding biosimilarity. Reducing IOV to 15% decreased required sample size substantially (~150), highlighting the impact of within-subject variability assumptions on design feasibility.
A worked example based on literature-derived perturbations reproduced the sample size of a previously successful crossover study (~112 subjects), supporting the realism of the framework [4,5].
CONCLUSIONS
This case study demonstrates how mechanistic PK/PD simulations can translate regulatory concepts of residual uncertainty into concrete biosimilar study design decisions. By isolating plausible sources of product-related variability, the workflow provides a transparent rationale for selecting a sensitive dose and determining the minimum sample size required for adequate power of a feasible biosimilarity study. In the context of evolving FDA guidance, such model-informed approaches enable pharmacometricians to move from supportive analysis toward directly informing biosimilarity development strategies. The openly available workflow facilitates practical adoption across products and programs.
References:
[1] FDA. Scientific Considerations in Demonstrating Biosimilarity to a Reference Product. 2015; Draft update Oct 2025.
[2] Li et al., Model-Based Approach to Selecting Pegfilgrastim Dose for Pharmacokinetic and Pharmacodynamic Similarity Studies in Biosimilar Development, 2022, DOI:10.1002/cpt.2722.
[3] Brekkan et al., Sensitivity of Pegfilgrastim Pharmacokinetic and Pharmacodynamic Parameters to Product Differences in Similarity Studies, 2019, DOI: 10.1208/s12248-019-0349-3.
[4] Brokx et al., A demonstration of analytical similarity comparing a proposed biosimilar pegfilgrastim and reference pegfilgrastim, 2017, DOI:10.1002/cpt.2722.
[5] Desai et al., Confirmation of Biosimilarity in a Pharmacokinetic/Pharmacodynamic Study in Healthy Volunteers for an Analytically Highly Similar Pegfilgrastim, 2016, DOI: 0.1002/cpdd.269.
Reference: PAGE 34 (2026) Abstr 11990 [www.page-meeting.org/?abstract=11990]
Poster: Methodology - Study Design